Homography superposition of television signals by RANSAC for commercial machine vision

The superposition of television signals is a topical issue that is considered in the computer vision systems when locating and recognizing objects, as well as to increase the image quality and decrease the volume of stored data. The article shows the algorithm of the homography superposition of television signals by RANSAC. The developed algorithm includes the following: interest point recognition, a point’s matching between input images by RANSAC and estimation of homography parameters. The optimal additional step is the superposition by the Lucas-Kanade method. This method allows us to estimate subpixel offset in the neighborhood area of interest point and decrease the error of superposition. The developed algorithm is intended to automate television signals superposition of computer vision systems, especially for the aviation equipment and security systems.